It was my understanding that this situation should not be possibile.
Currently, all updates affect only one of the two documents.
We're using version 0.17.8 in a cluster of two instances. Indeces were
initially created with version 0.11. The first copy of the document was
indexed before migrating to 0.17.8.
Is there anything I could investigate on to undestand what might have
happened?
It was my understanding that this situation should not be possibile.
Currently, all updates affect only one of the two documents.
We're using version 0.17.8 in a cluster of two instances. Indeces were
initially created with version 0.11. The first copy of the document was
indexed before migrating to 0.17.8.
Is there anything I could investigate on to undestand what might have
happened?
On Wednesday, November 23, 2011 4:10:30 PM UTC+1, kimchy wrote:
Are you using custom routing or parent when indexing? You can ask for
fields=_routing to get the routing value back as well.
I have never provided any routing*.* Adding fields=_routing doesn't have
any effect, that is: the query doesn't return any extra fields.
I also noticed another weird behavior. Our documents have a field called
"indexing_time". If I add it to the "fields" attribute, and repeat the same
query a few times, that field is not always included in the response. In
fact, it is included every other times. (It must have to do with our
cluster comprising two servers.)
On Wednesday, November 23, 2011 4:10:30 PM UTC+1, kimchy wrote:
Are you using custom routing or parent when indexing? You can ask for
fields=_routing to get the routing value back as well.
I have never provided any routing*.* Adding fields=_routing doesn't
have any effect, that is: the query doesn't return any extra fields.
I also noticed another weird behavior. Our documents have a field called
"indexing_time". If I add it to the "fields" attribute, and repeat the same
query a few times, that field is not always included in the response. In
fact, it is included every other times. (It must have to do with our
cluster comprising two servers.)
All attempts to recreate the problem have failed. Actually, documents are
very often reindexed, but the problem at issue seems to only
have occurred with that document. Replaying all operations on a fresh index
did not yield unexpected results.
I can, though, provide you with all the data files and the configuration.
It's just a few KB and does not include private information.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.